We present qualitative experiments on the CITIC RGB-D Dataset. Our goal is to show that only color information is not enough for background subtraction and it can be enhance with disparity images.
Figure 1: Crossing sequence. From left to right: (a) input, (b) variational, (c) phase, (d) SGBM, (e) ground truth, (f) DEOR-PCA variational, (g) DEOR-PCA phase, and (h) DEOR-PCA SGBM based foreground mask
Figure 2 : Lab door sequence. From left to right: (a) input, (b) variational, (c) phase, (d) SGBM, (e) ground truth, (f) DEOR-PCA variational, (g) DEOR-PCA phase, and (h) DEOR-PCA SGBM based foreground mask.